Characterization of Cocaine-Induced Block of Cardiac Sodium Channels

CORE Metadata, citation and similar papers at core.ac.uk

Provided by Elsevier - Publisher Connector

Characterization of cocaine-induced block of cardiac sodium channels

William J. Crumb, Jr. and Craig W. Clarkson Department of Pharmacology, School of Medicine, Tulane University, New Orleans, Louisiana 70112

ABSTRACT Recent evidence suggests reduce Na current in a use-dependent nisms of cocaine action on cardiac Na that cocaine can produce marked car- manner. The time course for block channels, we characterized its effects diac arrhythmias and sudden death. A development and recovery were char- using the guarded receptor model, possible mechanism for this effect is acterized. At 30 ,uM cocaine, two obtaining estimated Kd values of 328, slowing of impulse conduction due to phases of block development were 19, and 8,M for channels predomi- block of cardiac Na channels. We defined: a rapid phase (X = 5.7 ± 4.9 nantly in the rested, activated, and therefore investigated its effects on Na ms) and a slower phase (r = 2.3 ± 0.7 inactivated states. These data indicate channels in isolated guinea pig ventric- s). Recovery from block at - 140 mV that cocaine can block cardiac Na ular myocytes using the whole-cell vari- was also defined by two phases: (Tf = channels in a use-dependent manner ant of the patch clamp technique. 136 ± 61 ms, Tr = 8.5 ± 1.7 s) (n = 6). and provides a possible cellular expla- Cocaine (10-50 ,uM) was found to To further clarify the molecular mecha- nation for its cardiotoxic effects.

INTRODUCTION

Whereas the mechanism(s) for cocaine-induced mortality lated receptor model. One of the virtues of the model is remains poorly defined, it has been suggested that cocaine that it provides an analytical solution which allows the overdose may result in the production of cardiac arrhyth- calculation of rate constants for drug-channel interac- mias and sudden death (Benchimol et al., 1978; Wetli and tions. This model has been found to well describe the Wright, 1979; Nanji and Filipenko, 1984). Cocaine use effects of local anesthetics on action potential upstroke has become a major medical problem in the United velocity in heart tissue, but has not been adequately tested States, with overdose related deaths dramatically increas- using direct measurements on Na currents. Therefore, ing over the past decade (Mittleman and Wetli, 1984; our use of this model provides an additional test of the Pollin, 1985). Recent evidence suggests that blood levels model as well as the estimation of dissociation constants of cocaine achieved in individuals who died from cocaine for cocaine binding to channels in the rested, inactivated, overdose range from 5-90 ,uM (Finkle and McCloskey, or activated states. Our results provide experimental 1977; Mittelman and Wetli, 1984; Spiehler and Reed, evidence that cocaine has channel-blocking properties 1985). Because cocaine is well known to have potent local that make it a potent blocker of cardiac Na channels, and anesthetic properties in nerve, one possible mechanism for suggests a possible cellular explanation for the cardiotoxic cocaine cardiotoxicity is the production of cardiac con- effects of cocaine. A preliminary report of this work has duction disturbances resulting from block of cardiac been previously submitted in abstract form (Crumb and sodium (Na) channels. Because the cellular effects of Clarkson, 1989). cocaine on Na channels have not been well defined, we therefore investigated the effects of cocaine on Na channels in single cardiac myocytes isolated from guinea pigs METHODS using the whole-cell variant of the patch clamp technique (Hamill et al., 1981). We found cocaine to be a potent Isolation of cardiac myocytes blocker of cardiac Na channels. The method for isolation of ventricular myocytes used in this study is To further clarify the molecular mechanisms of cocaine similar to that described by Mitra and Morad (1985). Hearts from action on cardiac Na channels, we characterized its guinea pigs (200-300 g) were removed under pentobarbital anesthesia effects using the guarded receptor model (Starmer et al., and mounted by the aorta onto the cannula of a Langendorff perfusion apparatus. Hearts were initially perfused for 5 min at 10-15 ml/min 1984; Starmer and Grant, 1985; Starmer et al., 1986; with a calcium-free bicarbonate buffered solution at 370C followed by Starmer, 1987). Starmer et al. (1984) proposed the an 8-min perfusion with the same solution supplemented with 2 mg/ml guarded receptor model as a simplification of the modu- collagenase (284 U/mg; Worthington Type II; Worthington Biochemi-

Biophys. J. Biophysical Society 0006-3495/90/03/589/11 $2.00 589 Volume 57 March 1990 589-599 cals, Freehold, NJ) and 0.28 mg/ml protease (5.6 U/mg; type XIV; Evaluation of methods Sigma Chemical Co., St. Louis, MO). The ventricles were then cut free, placed in an enzyme-free solution containing 0.2 mM CaC12, minced To accurately study drug effects on Na currents, other ionic currents with scissors, and gently stirred at 370C for 5 min. Cells were harvested must be insignificant, and voltage clamp errors related to uncompen- from the suspension by filtration through a 200-gm nylon mesh, and the sated series resistance and voltage inhomogeneity reduced to within filtrate was then centrifuged at 200-300 rpm for 1.5 min to form a acceptable levels. In these experiments, Na currents were isolated from pellet. The supernatant was suctioned off and the pellet was resuspended Ca ++ and K + currents by replacing all K + with Cs +, and eliminating in enzyme-free solution supplemented with I mM CaCl2 and I mg/ml Ca + + currents by blocking Ca + + channels with internal F ions. Under bovine serum albumin (fraction V; Sigma Chemical Co.). This suspen- these conditions steady-state (leak) current at the test potential (-20 sion was also centrifuged to form a pellet, and this pellet was mV) was <0.1 nA, or <1-2% of peak Na current. Correction for leak resuspended in minimal essential medium supplemented with Earle's current was not made. Values of cell capacitance were estimated from salts, L-glutamine (292 Mg/ml), 3% horse serum, 0.1 mg/ml streptomy- integration of capacitive transients, and had a mean value of 66 ± 4 pF cin, and 100 U/ml penicillin (Gibco Laboratories, Grand Island, NY). (mean ± SD, n 13). Following rupture of the membrane seal, the This final suspension was poured into culture dishes and stored at 370C mean total series resistance (Rs) for the pathway between pipette and in a CO 2 incubator until the cells were used in experiments (within 30 h cell membrane was 1.3 ± 0.1 Mg (mean ± SD, n 13) under the after isolation). The cells used in experiments were of characteristic experimental conditions used in this study. It was possible to electroni- morphology (rod-shaped with well-defined striations) and lacked any cally compensate for 60-80% of this series resistance in these cells. visible blebs on the surface. Experiments were performed in cells where the estimated voltage drop across the uncompensated series resistance was <5 mV. Data analysis and mathematical Solutions modeling The solution used for perfusing the heart and cell isolation consisted of An IBM-PC/AT using programs written in the ASYST language Joklik modified minimum essential medium (Gibco Laboratories) (Asyst Software Technologies, Inc., Rochester, NY) was used for curve adjusted to a pH of 7.2 with NaOH, gassed with 95% 02:5% CO2, and fitting and mathematical modeling. A nonlinear least-squares algorithm warmed to 370C. When electrical recordings were made from isolated using the Gauss-Newton method was used to fit exponential functions to cells, the bath was perfused with an "external" solution that consisted of experimental data. (in millimolars): 115 tetramethylammonium chloride, 25 NaCI, 5 CsCl, To further define the effects of cocaine on cardiac Na channels, 1.8 CaC12, 1.2 MgCl2, 20 Hepes, 11 glucose; adjusted to a pH of 7.3 apparent rates of drug binding and unbinding to rested, open, and with tetramethylammonium hydroxide. Glass pipettes (electrodes) were inactivated Na channels were calculated using analytical equations filled with an "internal" solution that consisted of (in millimolars): 125 based upon the guarded receptor model (Starmer et al., 1984; Starmer CsF, 20 CsCl, 10 NaF, 10 bis-Tris propane, 5 EGTA; adjusted to a pH and Grant, 1985; Starmer, 1987). According to the model, tonic and of 7.2 with CsOH. The cocaine HCI was purchased from Sigma use-dependent block of Na channels observed during a train of depolar- Chemical Co. izing pulses results from drug interaction with a receptor site having diffusion pathways that are guarded by the position of channel activation and inactivation gates (Fig. 7). When all activation gates are closed, drugs cannot bind to the channel receptor since these gates prevent Voltage clamp recording access to the receptor site via both hydrophilic and hydrophobic pathways. Therefore at very negative potentials, where the fraction of At the beginning of each experiment, an aliquot of cells was placed into channels having open activation gates is small, only a relatively small a shallow tissue bath mounted atop an inverted microscope. Sodium tonic block is observed at low drug concentrations. When channels open, currents from the isolated ventricular myocytes were measured using drug can bind and unbind from the receptor over both hydrophilic and the whole-cell variant of the patch clamp method (Hamill et al., 1981). hydrophobic pathways. Once the inactivation gate has closed and the The glass pipettes used were pulled in two stages then fire polished until hydrophilic pathway is obstructed, drug may continue to bind and tip openings of 3-4 ym were obtained. Tip resistances were 0.3-0.5 MQl unbind from the channel receptor via the hydrophobic (membrane) when the pipettes were filled with the internal solution. Pipette tips were pathway. In a channel occluded by a charged drug, closing of either the positioned medially along the long axis ofthe cell, and a gigohm seal was activation or inactivation gates prevents drug dissociation until either formed between the pipette tip and cell membrane. The membrane the channel reopens, or as for a tertiary amine, the drug is converted to patch under the pipette tip was then ruptured by applying a suction an uncharged (membrane permeant) form. Uncharged drugs are not transient, creating a continuity between the internal (pipette) solution trapped in rested or inactivated channels since they can diffuse into the and the cell interior. The bath temperature was measured by a thermis- lipid bilayer at any time. As a consequence of receptor guarding and tor placed near the cell under study and was maintained at 1 60C ± 0.50C drug trapping by channel gates, a pulse-by-pulse accumulation of with a thermoelectric device (model 806-7243-01; Cambion/Midland channel block (use-dependent block) is observed when channels are Ross, Cambridge, MA). repetitively cycled between hyperpolarized and depolarized potentials. An Axopatch 1-B amplifier (Axon Instruments, Burlingame, CA) To estimate the apparent rate constants defining drug interaction was used for whole-cell voltage clamping. Creation of voltage clamp with channels once they become inactivated, it was assumed that the pulses and data acquisition were controlled by an IBM PC/AT slow component of time-dependent block characterized using a two- computer running pClamp software (Axon Instruments). The computer pulse protocol (Fig. 3 A) results from drug binding to a continuously was interfaced to the amplifier by an 80-kHz Labmaster board (Scien- accessible receptor (i.e., via the hydrophobic membrane pathway) as tific Solutions, Inc., Solon, OH). Sodium currents were digitized at described by sample intervals of 20-200 js. Cells were maintained at a holding potential of - 140 mV between pulse protocols to fully remove inactiva- [D] + U B, tion. lj

590 Biophysical Journal Volume 57 March 1990 where U and B represent the fraction of unblocked and blocked To estimate the rate constants for drug binding and unbinding to rested channels, [DI is the drug concentration, and k j, I are apparent rates of and open-channel states, estimates of the parameters X, bo, and b, were drug binding and unbinding. According to this model, changes in the first obtained by fits of Eq. 5 to peak Na currents obtained during pulse proportion of blocked channels observed during a long pulse is an trains at five to six different interpulse intervals. The estimated value of exponential process defined by the equations: bo was taken as an estimate of the equilibrium level of block at the holding potential (R.,). Estimates of X, and Xo were then computed -1 to X vs. tr = Xi = (kj[D] + Ii); = (1) according Eq. 6 by linear regression of (Xr slope, Xo [X-intercept - XAt1J/to). For analysis, to was assumed to be equal to 1 1 = + I,) or I1,= = (2) ki[D]/(ki[D] ki[D]Ti ms (i.e., similar to reported values of the mean open time for single ki = I/(Ti[D]) (3) cardiac Na channels) (Fozzard et al., 1987; Kunze et al., 1985) and t, was assumed to be equal to the duration of the test pulse at - 20 mV - Ii = (1/Ti) - (ki[D]), (4) to. The equilibrium level of open channel block at the test potential (0_.) was estimated from Eq. 11 by linear regression of bM vs. XI (O,, where k i and I reflect the apparent rates of drug binding and unbinding slope). to inactivated channels, Ti is the time constant for drug binding to From calculated values of Xr, R_,, and 0,,, the apparent forward (kr, inactivated channels, and I_ is the steady-state fraction of inactivated- k.) and reverse (Ir, l1) rate constants defining drug interaction with blocked channels. These rate constants were calculated from estimates rested and open channels were then calculated using the following of the time constant (T i) and steady-state level (I -.) of the slow phase of equations (Starmer, 1987) block development defined using a two-pulse protocol (Fig. 3 A). The rate constants defining drug interaction with open and rested (13) kr = XrR.I[D] channel states were estimated from rate train data using an analysis (14) scheme similar to that previously used by Starmer and co-workers ko= XO..I/[D] (15) (Starmer and Grant, 1985; Starmer et al., 1986; Starmer, 1987). lr = X,(l -R) However, because the experimental data indicated that cocaine can (16) = U(l - 0..). interact with both open and inactivated channels, the analytical equations of Starmer et al. were modified to incorporate the presence of three The guarded receptor model predicts that these state-dependent rate (vs. two) significant channel states (rested, open, and inactivated). The constants can theoretically be broken down further into guard and trap time course of use-dependent block onset during a train of repetitive functions and state-independent binding and unbinding rates once the pulses was described in terms of a recurrence relation (Starmer and model for channel gating and deprotonation/protonation rates for Grant, 1985): bound drug are known (Starmer and Grant, 1985; Starmer and Court- ney, 1986). However, in this study, only the apparent "state-dependent" bM + - (5) bn= (bo b,)e-"x, rate constants were defined because neither the proper gating model for where bn is fraction of channels blocked during the nth pulse, b. is the cardiac Na channels, nor the conversion rates between charged and uncharged cocaine within the channel were known. fraction of channels initially blocked (e.g., after a 1-min rest at the holding potential), b. is the "steady-state" fraction of channels blocked after many pulses, n is the pulse number, and X is the blocking rate constant. For a three-state model, the blocking rate constant X can be Statistical analysis further defined by Paired and unpaired two-tailed Student's t tests were used to evaluate significance between two groups of data. Comparisons between more X = Xr tr + Xo to + Xi ti, (6) than two groups was accomplished by one-way analysis of variance were whereXrA, X0, andXi are the rates of drug interaction with channels when (ANOVA). Data analyzed by linear regression evaluated by they are predominantly in rested, open, or inactivated states during the calculation of the Pearson product-moment correlation coefficient (r) pulse protocol, t, is the interpulse (recovery) interval, to is the channel and a t test for significance of linear trend. If values of P < 0.05 were the differences were are mean open time, and ti is the time interval channels spend in the obtained, considered significant. Results inactivated state (Starmer and Grant, 1985). The steady-state level of expressed as mean ± SD. channel block (b,,) may also be defined by (Kojima and Ban, 1988; their Model I): RESULTS b, = O..X, + I..X2 + R..X3, (7) Tonic and block where use-dependent (8) Cocaine reduced cardiac Na current in a manner similar X= (1 - e-Aoto)e-(Aii+V,)I(l e_ (9) to many other local anesthetics. This reduction in current X2= ( - e-Xiti)e-rtr/(l - e-X) (10) amplitude could be divided into two components: tonic X3 = (1 -e-rtr)/(1 -e-A). and use-dependent block. As indicated in Fig. 1 B and C, upon of a train of at 5 Hz Eq. 7 may also be simplified as application depolarizing pulses after a long rest, there was no noticeable change in Na

= bu O. XI + 5, (11) current amplitude in the absence of drug. After a 10-20 exposure to or 50 there was where min either 10, 30, ,uM cocaine, a reduction in Na current amplitude during the first pulse 6 = I..X2 + R.X3. (12) after a long rest (tonic block) as well as an additional

Crumb and Clarkson Cocaine-induced Block of Cardiac Sodium Channels 591 20 ms A condition test A -20 mV pEVcl lOms -20 mV

s -140 rnV n=30 -140 mV - 0m -2ms 3 Hz B n=15 B (c Control z 1 -h_ - Control 5 MCocaine ,. VIJ. _u_ _ a a R _IJACMI~lrw"U-G.-O1

<1 ~~~~~~5Hz Cucos 0.8 ,2 ms 0.6 Cocaine z lw 0 0 0 0 C ,Control 0.4 peak GNa

0 0.2 f.'UN.dt 0.0 -140 -100 -60 -20 20 60 z a) Vc (mV)

a1) c: FIGURE 2 Voltage-dependence of use-dependent block produced by 30 uM cocaine. (A) Pulse protocol. Use-dependent block was produced by a train of 15 pulses of 2 ms duration at 3 Hz to selected conditioning - 0 1 2 3 4 5 potentials (Vs). The amplitude of the peak Na current available after the conditioning train was determined with a test pulse after a recovery Stimulation Rate (Hz) interval of 300 ms. (B) Shows the relationship between peak Na current and the conditioning potential (Vj). Peak current was normalized to its - is control, closed circle is FIGURE I Tonic and use-dependent inhibition of the peak sodium (Na) value after a long rest at 140 mV. Open circle current by cocaine in guinea pig ventricular myocytes. (A) Pulse 30 MM cocaine. The relation between membrane potential and GN. for protocol. A train of depolarizing pulses to -20 mV was applied at the same cell is also shown for comparison. GN. was calculated using different stimulation rates after a I min rest at the holding potential both peak Na current (closed squares) and the integral of Na current (-140 mV). (B) Superimposed records of Na current obtained during a over a 2-ms interval (open squares). train of pulses applied at 5 Hz before and during exposure to 50 OM cocaine. For the controls, currents for only the 1st and 30th pulses are shown. There was no noticeable decrease in peak current during Voltage dependence repetitive pulsing under control conditions, whereas in the presence of drug there was both an initial (tonic) block observed on the first pulse, as The relationship between membrane potential and Na well as a use-dependent block that approached a steady-state within 30 channel block was determined by using the pulse protocol pulses. (C) Shows the mean level of peak Na current after 30 pulses at selected stimulation rates under control conditions (open circles), in the shown in Fig. 2 A. A conditioning train consisting of 15 presence of 10 MM cocaine (closed squares), 30 MM cocaine (open pulses of 2-ms duration was applied to various voltages to diamonds), and 50 MAM cocaine (closed circles). Peak Na current produce a steady-state level of block. After a 300-ms amplitudes have been normalized to the amplitude ofcontrol Na current recovery interval, a test pulse was given to determine the after a >10-s rest. Data values represent mean ± SD (n - 4-6). amount of block produced by the conditioning train. The 300-ms recovery interval was long enough to allow unblocked channels to recover from inactivation, but short enough to allow minimal recovery from cocaine- reduction in current amplitude which became progres- induced block. sively larger with each successive pulse (use-dependent In the absence of drug, the conditioning train had little block), reaching a steady state after 15-30 pulses. This effect on the amplitude of the Na current recorded during steady-state level of block was frequency dependent. the test pulse (Fig. 2 B). However the addition of either Mean levels of both tonic and use-dependent block are 10, 30, or 50 ,uM cocaine resulted in a voltage-dependent shown in Fig. I C. In the absence of drug, there was no channel block. As indicated by the smooth curve in Fig. 2 noticeable effect of pulsing on Na current amplitude as B the voltage dependence of block could be well described stimulation rate was increased to 5 Hz. In contrast, with by an equation used to describe open channel block by the addition of 10, 30, or 50 ,M cocaine there was up to a local anesthetics in nerve (Cahalan, 1978; Cahalan and 10% reduction in Na current amplitude upon the first Almers, 1979): pulse. In addition there was a significant use-dependent block which developed upon repetitive stimulation at INa (drug)/1Na (contrl) rates above 0.2 Hz (P < 0.05). = (1 - B)/(1 + exp [Vc- Vmid/S]) + B, (15)

592 Biophysical Journal Volume 57 March 1990 level of Na current A where B is the asymptotic remaining 1.0 Control after conditioning trains to positive potentials, V, is the membrane potential during the conditioning train, Vid is 0.8 Condition Test \ -20 mV 10 ms 20 mV the membrane potential at half-maximal block, and S is 0 0.6-\ r block. Estimated mean mr:X 0.3S -140mV the slope of the voltage dependent a) 0.4 values of these variables obtained from least squares fits I- - to Eq. 15 using data obtained in the presence of 10-50 ,uM 0.2 30 pM Cocaine cocaine (n = 5) were -51.1 ± 6.9 mV for the midpoint - , ± l' for voltage-dependent block (Vmid) and 13.7 3.1 mV for 0 22, 4,4 6 8 10 the slope factor for block (S). Condition Pulse Duration (s) To determine whether the voltage dependence of chan- B m 0 nel block was similar to channel opening, channel block 1.0 PC-I, was compared with Na channel conductance (GNJ). GN. 0.8 o Contitrol was calculated from peak INa, and from integrations of 8 iM Lidocaine the 2 ms or over the entire 50 ms of 0.6 INa during initial 12 *50lp1 M Cocaine depolarizing pulses to potentials from -90 to + 10 mV. 0 0.4 Condiion Test

The 2- and 50-ms integrals were selected because they 5 sec - 0.2 -20 mvl -2Cto mv closely resemble either the conditioning pulse duration -140 mVn 4H -dt_ (Fig. 2 A) or an integral of INa over a time period during 0 -- which the Na current is almost completely inactivated 0 8 16 24 32 40 (50 ms). Peak GN5 fits obtained using a Boltzman equa- Recovery Time (s) tion of the form GN. = G,,,/[1 + exp (Vmjid- Vm)/SI had a mean half-maximal GN, value (Vmid) at -55.6 ± FIGURE 3 (A) Biphasic time course of block development at -20 mV. 7.1 mV (n = 5) and a mean slope factor (S) of 4.4 ± 1.7 (Inset) Pulse protocol. Block development was determined by a single mV (n = 5). Thus although both channel block and peak test pulse applied 300 ms after a conditioning pulse ofvariable duration. GNa were half maximal at similar potentials, the slope The interval between each trial of the pulse protocol was 1 min. Panel A shows the time course of block development at - 20 mV. Under control factor for peak GN5 was significantly smaller (P < 0.001) conditions (open circles) there was little change in peak current than that for channel block, indicating that channel amplitude following conditioning pulses of up to 10 s duration. In the opening had a steeper voltage dependence than channel presence of 30 MM cocaine, peak Na current was inhibited in two block. The slope factors for time integrals of GN. were distinct phases. Test currents have been normalized to their value after a also significantly smaller than that of channel block [for a long rest in the absence of a prepulse (I.). (B) Time course of recovery from block at the holding potential (-140 mV). (Inset) Pulse protocol. 2-ms integral: S = 6.1 ± 0.7 (P < 0.001), -43.3 + Vmid A steady-state level of block was produced by a conditioning pulse of 5 s 6.6; and for a 50-ms integral: S = 2.5 2.6 (P < 0.01), duration to -20 mV. The kinetics of recovery from block were then Vmid = -63.5 ± 5.3](n = 5). defined by measuring the fraction of peak Na current (IT<") available with a test pulse elicited after a variable recovery interval. Panel B shows the time course of recovery of peak Na current under control conditions Block development (open circles), 50 uM cocaine (closed circles), and for comparison 100 uM lidocaine (closed squares). The test currents have been normalized To characterize the ability of cocaine to bind to Na to their steady-state rested values (I..). In the presence of 50 yM channels during a depolarizing pulse a two-pulse protocol cocaine, recovery from block consisted of two phases and was well fit by was employed to determine the time course of block the sum of two exponentials. The slow phase of recovery with 50 /M development. As illustrated in Fig. 3 A, the protocol cocaine had a time constant (r,) of 8.5 ± 1.7 s as opposed to that of 100 MM lidocaine where r, is 1.0 ± 0.4 s. consisted of a conditioning pulse of variable duration at

- 20 mV followed by a test pulse after a 300-ms recovery interval. The recovery interval was long enough for most drug-free channels to recover from inactivation, but short equation: Afe-'t'f + A5e"'It + A.. The amplitudes and enough to allow only a minimal (<4%) amount of recov- time constants for block development in the presence of ery from channel block. In the absence of drug, test 30 gM cocaine were: Ar = 0.14 + 0.08, Tf = 5.68 ± 4.92 current amplitude was not substantially depressed even ms, A, = 0.64 ± 0.06, 7r =2.29 + 0.68 s, and A. = 0.23 + with conditioning pulse durations of up to 10 s. However, 0.03 (n = 5); and those for 50 MM cocaine (not shown in with the addition of 30 MM cocaine, test current ampli- Fig. 3 A) were: Af = 0.07 ± 0.01, rf 2.88 ± 1.47 ms, tude progressively decreased as the conditioning pulse As = 0.79 ± 0.01, r, = 1.46 ± 0.61 s, and A., = 0.13 + duration was increased. The time course of block develop- 0.02 (n = 4). The time constants for both phases and the ment consisted of two distinguishable phases (fast and amplitude of the fast phase were not significantly dif- slow) which could be well approximated by the following ferent at either concentration, however, the amplitudes of

Crumb and Clarkson Cocaine-induced Block of Cardiac Sodium Channels 593 the slow phase and the steady-state level of block were TABLE 1 State-dependent rate constants for cocaine HCI* significantly different (P < 0.001). Parameter Value

Recovery from block kr(M-'S-') 274 ± 217 The time course of recovery from block was characterized lr(s') 0.1 ± 0.05 using the two-pulse protocol depicted in Fig. 3 B. A KdA(UM) 328 ± 156 ko(M-'s-') 1.13 x 106 ± 2.5 x 103 steady-state level of block was produced by a single UO(S`') 20 ± 10.2 conditioning pulse to - 20 mV for 5 s. The time course of Kd,(AM) 18.5 ± 10.2 recovery from block was then determined by a test pulse kj(M-'s-') 1.2 x 104 ± 5.0 x 103 after a variable recovery interval. In the absence of drug, li(s-') 0.1 ± 0.05 Kdi(JUM) 7.9 ± 1.4 recovery from inactivation at - 140 mV was described by a double exponential function having a large fast compo- *Data are grouped mean ± SD (n - 10), derived from rate constants for nent with an amplitude (Af) of 0.87 ± 0.05 and a time 30 and 50 ,uM cocaine. Rate constants are apparent binding and constant (rf,) of 13 ± 4.0 ms, and a small slow component unbinding rates to channels rested at -140 mV, activated at -20 mV, and inactivated at - 20 mV. No attempt was made to factor out receptor with an amplitude of 0.18 ± 0.05 and a time constant (Aj) guard and trap functions (see Methods). State-dependent dissociation (Ts) of 246 ± 163 ms (n = 8). In the presence of 50 ,uM constants (Kd) are the ratio of I/k. cocaine, recovery of the Na current after the conditioning pulse displayed two phases (Fig. 3 B): a small fast phase (Af = 0.13 ± 0.03, Tf = 136 ± 61 ms), which probably reflects a combination of recovery of drug-free channels level of block (tonic + time-dependent) at -20 mV (I.), as well as rapid unblocking of drug from open channels; and the time constant for the slow phase of channel block obtained using a two pulse protocol (Fig. 3 A). Mean and a large slow phase (A, = 0.85 + 0.03, Tr 8.5 ± 1.7 s) (n = 6), which reflects recovery of drug-associated chan- values of ki and l are shown in Table 1. To calculate the nels from a rested state. Alteration of the recovery apparent rates of drug binding to rested and open (activated) data from train potential over a 40-mV range (from - 120 to - 150 mV) channels, pulse protocols (Fig. 5 did not significantly alter the time constant for recovery A) were analyzed using a parameter estimation procedure from block (Fig. 4). based on the guarded receptor model (see Methods). A nonlinear least squares algorithm was used to obtain fits Mathematical modeling of cocaine block A 20 ms -20 mV The apparent rate constants defining drug binding (k1) and unbinding (1l) from inactivated channels were calcu- -140 mV lated using Eqs. 1-4 from estimates of the steady-state B n=30

100r 30 MM Cocaine 0 10 a a z ° 1.0

cc 01 0.1 -150 -140 -130 -120 Vm (mv)

Pulse Number FIGURE 4 Voltage dependence of recovery from block by 30 gM cocaine. A steady-state level of block was produced with a 5 s pulse to -20 mV. The voltage dependence of cocaine unblocking was defined by FIGURE 5 Guarded receptor analysis of use-dependent block of cardiac measuring the slow component of recovery from block at different Na channels by cocaine. (A) Pulse protocol. A train of depolarizing holding potentials (Vm) (-150 mV - 7.4 ± 4.1 s, -140 mV - 8.2 ± 2.3 pulses to - 20 mV was applied at different stimulation rates following a s, - 130 mV - 8.4 ± 2.4 s, and - 120 mV - 9.3 ± 1.1 s) (n - 3-8). The 1-min rest at the holding potential (-140 mV). (B) Peak Na currents equation of best fit was Tr 8.4 e (0.21vm) Values are given as mean ± recorded during pulse trains at different rates were fit to Eq. 5 to SD. estimate uptake rates (X) and steady-state amplitudes of block (b,).

594 Biophysical Journal Volume 57 March 1990 of Eq. 5 to peak Na current values recorded during trains A 0.60 of 20 ms pulses to -20 mV (Fig. 5 B). From these fits, ^ 0.48: values of rate tonic block (bo), blocking (X), and steady- X 0.36: state level of use-dependent block (b,) were estimated for o0.24 r= 1.00 0. (P<0.01) 5-6 different interpulse intervals (tr). A linear regression 0.12 0 of blocking rate (X) against recovery interval (t,) was then 0 1 2 3 4 5 performed using Eq. 6 to calculate values of X, and X. Recovery Time tr (s) (Xr = slope, Xo = [X-intercept Xtj]/t.)(Fig. 6 A). Val- B ues of the blocking parameter X, were then computed 0.80- using Eq. 8, and a linear regression of steady-state block 0.64- (b,,) against X, was performed using Eq. 11 to estimate 0.48-- the equilibrium levels of channel block at -20 mV 0.32 r = 1.00 (slope = O.) (Fig. 6 B). The relationships between both X 0.16- / (p<0.01) vs. and vs. were to a t, b. X1 found be well described by 0 simple linear function (r > 0.9, P < 0.05) (Fig. 6 A and 0 0.16 0.32 0.48 0.64 0.80 B), consistent with the theoretical model. These estimated X1 model parameters were then used to calculate the apparent binding and unbinding rates according to Eqs. 13-16. FIGURE 6 Mathematical modeling of channel block by cocaine accord- Mean values for k, I,r k., and 1 are given in Table 1. ing to the guarded receptor model. (A) Shows a fit of the relationship Since the estimates of ko and 10 were based upon an between blocking rate (A) and the interpulse recovery time (t,) according to Eq. 6. (B) Shows a fit of the relationship between b, and assumed value for to (1 ms), assumption of a different X, according to Eq. 7. The observed curves are in agreement with the access interval for calculation of Xo will result in recipro- theoretical predictions of an exponential relationship between Na cur- cal changes in estimated values of ko and 10 from Eqs. 14 rent amplitude and pulse number (Fig. 5 B), as well as linear relation- and 16 (e.g., assuming a value of to = 2 ms results in ships between A vs. t, and b. vs. X,. values of ko and IO half those appearing in Table 1).

hypothesis that this component of block was due to block of open Na channels in cardiac myocytes. DISCUSSION A recent single channel study of cocaine's effects on Block development neuronal Na channels in planar lipid bilayers (Wang, 1988) indicates that cocaine can block open neuronal Na The time course of cocaine block of cardiac Na channels channels in a voltage-dependent manner. Using a pulse during a depolarizing pulse was characterized using a protocol (Fig. 2 A) similar to that previously used to two-pulse protocol (Fig. 3 A). As with other local anes- define the voltage dependence of local anesthetic block in thetics (Clarkson et al., 1988), cocaine block consisted of nerve (Clarkson et al., 1988), we found that the rapid two definable phases: a rapid phase (Tf = 3-6 ms) and a component of channel block had a voltage dependence slow phase (r, = 1.5-2.3 s), indicating that cocaine exhib- that was similar to that of channel opening: both peak GNa its a relatively high affinity for multiple channel states and channel block developed and reached an asymptote occupied during a depolarizing pulse. During the first few over the same voltage range and had voltage midpoints milliseconds of a suprathreshold depolarizing pulse, Na that were not significantly different (-55.6 vs. -51.1 channels fluctuate through multiple conformational mV) (P > 0.3). This differs from the voltage dependence states, thus making it difficult to accurately identify the of inactivation which under the conditions of this experi- channel states occupied during the rapid phase of block. ment develops and saturates over the voltage range of Previous studies on other local anesthetics have attributed - 120 to -60 mV (Follmer et al., 1987; Clarkson et al., a rapid phase of channel block to the charged (cationic) 1988). Although the voltage-dependence of the fast com- form of the drug binding to the open-state of the Na ponent of block developed and saturated over a similar channel (Strichartz, 1973; Hille, 1977; Hondeghem and voltage range as that of channel opening, the two were not Katzung, 1977; Starmer et al., 1984). Under the condi- identical in that channel block was consistently less steep tions of these experiments (pH = 7.2-7.3, tempera- than channel opening. The explanation for this difference ture = 15-160C), cocaine (pKa = 8.6) should exist is not clear but could possibly be due to competitive largely (95-96%) in its charged form. The fact that interactions between cocaine and monovalent cations cocaine exists primarily in the charged (cationic) form (e.g., Na +) at a common binding site, as recently reported and has a component of block development that overlaps in lipid bilayers (Wang, 1988). Wang found that as that of channel opening lead us to further test the external Na+ was increased from 100 to 300 mM, there

Crumb and Clarkson Cocaine-induced Block of Cardiac Sodium Channels 595 was an increase in drug dissociation rate resulting in a unbinding include: (a) trapping of drug in an inactivated- sixfold increase in the calculated Kd value for drug blocked state due to a drug-induced hyperpolarizing shift interaction with open channels. It therefore seems reason- in the voltage dependence of the inactivation gating able to speculate that an increase in the external Na + mechanism (Hille, 1977; Hondeghem and Katzung, from that used in this study (25 mM) to more physiologic 1977; Schwarz et al., 1977; Hondeghem and Katzung, concentrations might similarly increase the cocaine disso- 1980; Gintant and Hoffman, 1984) or (b) trapping of ciation rate and the estimated Kd value for cocaine charged cocaine molecules behind closed activation gates interaction with open channels in the heart as well. (Strichartz, 1973; Courtney, 1975; Hille, 1977; Yeh and The voltage-dependence of cocaine block was not ade- Tanguy, 1985; Starmer et al., 1986). quately fit by a previously proposed model describing the We believe that the slow component of drug unbinding voltage dependence of channel block to the effect of cannot be explained in terms of drug trapping in an changes in the transmembrane voltage gradient on the inactivated-blocked state at the holding potentials stud- passive movement of drug to a binding site within the ied. For example, using the equation of Bean et al. (1983) channel (Strichartz, 1973; Cahalan, 1978). This type of (AVh,. = kln[Kdl/KdR], where k is the slope factor of the model predicts the slope parameter (S) which defines the inactivation curve), one arrives at a value of - -22 mV steepness of voltage-dependent block should be equal to for the cocaine-induced voltage shift of inactivation (as-

RT/zbF, where F, R, and T have their usual meanings, z suming Kdl and KdR values shown in Table 1, k - 6 mV) i's the drug molecule's valence, and a is the equivalent (Clarkson et al., 1988). The midpoint of the inactivation electrical distance of the binding site across the mem- (h.,) curve was estimated to be -93 ± 3 mV (n = 16). brane field (0-1) from inside to out (Cahalan, 1978). Our Therefore, clamping the membrane to potentials negative experimental conditions were such that RT/F = 25 mV to - 120 mV would be expected to completely surmount a and z = 1. Therefore the minimal value of S predicted by drug-induced shift in inactivation of this magnitude. this model is 25 (when 6 = 1). This contrasts with the Hyperpolarization to potentials that surmount a voltage mean slope values of 13.7 3.1 mV for cocaine block. shift should result in both rapid recovery of Na current at These results indicate that the observed dependence of the the holding potential as well as an abolishment of use- fast component of cocaine block on membrane voltage dependent block when drug trapping in an inactivated- cannot be attributed to movement of charged drug across blocked state is the rate-limiting step regulating channel the membrane. An alternative explanation is that the unblocking (Hondeghem and Katzung, 1980). This was voltage dependence of channel block results indirectly not seen experimentally (Fig. 1), even when the mem- from the voltage dependent availability of a channel state brane was clamped to - 160 mV. Thus the trapping of to which drug binds with relatively high affinity. The channels in an inactivated-blocked state cannot explain similarity between the voltage-dependence of channel the slow rate of cocaine unbinding at hyperpolarized block and peak GNa suggests that the high-affinity state potentials (- 120 to - 160 mV). responsible for the rapid component of block is either the An alternative explanation for slow unblocking at open state, or a closely associated "activated" channel hyperpolarized potentials is trapping of charged drug state. behind closed activation gates (Starmer and Grant, 1985; The slow component of channel block can be most Yeh and Tanguy, 1985; Starmer et al., 1986). At very easily explained by drug binding to inactivated channels. negative potentials (i.e., negative to - 120 mV), the This explanation seems plausible because virtually all Na activation gates in most channels are closed. Closed channels become inactivated within 100 ms at -20 mV. activation gates may regulate the rate of drug efflux and Other tertiary amines have also been shown to produce a influx to the channel receptor (Starmer and Grant, 1985; slow phase of channel block in cardiac tissue (Bean et al., Yeh and Tanguy, 1985; Starmer et al., 1986). However, 1983; Clarkson et al., 1988). explaining slow unbinding in terms of activation gate trapping is only applicable if the drug molecule exists Recovery primarily in the charged form, and/or cannot readily escape from the channel lumen by diffusion into the lipid As indicated in Fig. 3 B, recovery of Na current in the membrane. Although cocaine (pK = 8.6) exists primarily presence of a high concentration of cocaine exhibits two in the cationic form at the pH used in these experiments phases: (a) a small rapid phase which may be due to (7.2-7.3), because it is a tertiary amine, it is expected to either recovery of channels which were never blocked or undergo periodic deprotonation to an uncharged (lipid rapid open-channel unblocking of cocaine-blocked chan- soluble) form. This suggests that the rate of drug efflux nels during the test pulse and, (b) a large slow phase. Two from closed channels at the holding potential could be possible explanations for the slow phase of cocaine rate-limited by the kinetics of drug deprotonation to the

596 596 Biophysical Journal Volume 57 March 1990 lipid soluble form rather than the rate of infrequent to adequately simulate the use-dependent effect of channel openings. Assuming a protonation rate (kp) of cocaine on cardiac Na channels. As predicted by the 5 x 108 M -'s -' (Schwarz et al., 1977), the lifetime of the model, the time course of use-dependent block was well cationic form of cocaine (1 /lp) is predicted from (kp/lp = described by a single exponential function (Fig. 5 B), and 1IOP') (Schwarz et al., 1977; Starmer and Courtney, linear relationships were observed between both the 1986) to be =0.8 s. Because the time constant for channel uptake rate (X) and the interpulse recovery time (tr), as unblocking observed experimentally was an order of well as the stimulus parameter XI and steady-state block magnitude larger (.8 s) it is apparent that the deprotona- (b,) (r > 0.9, P < 0.01) (Fig. 6, A and B). In addition, tion rate predicted for cocaine in aqueous solution cannot both the kinetics and amplitude of block by cocaine were solely explain the slow rate of drug unbinding, and other closely described by "state-dependent" rate constants variables may be involved. A previous study by Woodhull calculated from the guarded receptor parameter estima- (1973) has suggested that hyperpolarized potentials can tion procedure (Table 2). attract H + from the extracellular solution into the chan- Upon inspection of Table 1, it is evident that the model nel lumen. An increased proton concentration within the rate constants defining drug binding (k0) and unbinding channel would be predicted to increase the probability of (is) from activated channels are two to four orders of a proton-cocaine collision and the probability of drug magnitude larger than those for drug binding and unbind- protonation. An increased probability of drug protonation ing from rested or inactivated channels at concentrations would be expected to prolong the existence of the cationic of 30 and 50 ,uM cocaine. As illustrated in Fig. 7, binding species within the channel. In addition, the pKa ofcocaine and unbinding may be regulated by the position of the within the channel lumen could be different from that of activation and inactivation gates (Starmer et al., 1984; cocaine within the bulk solution (Starmer and Courtney, Starmer and Grant, 1985; Starmer et al., 1986; Starmer, 1986). Taking this into consideration, the rate of unbind- 1987), thus accounting for these differences in state- ing can be more closely approximated if a pKa of 9.6 (vs. dependent rate constants. 8.6) is assumed for cocaine molecules within the channel Although the guarded receptor equations can ade- lumen. Therefore, although the slow rate of cocaine quately simulate most of our experimental data, this unbinding from rested channels seems consistent with observation does not indicate that other models could not activation gate trapping, it is possible that changes in describe the data equally well. Nevertheless, the close local pH within the channel and/or changes of drug pKa agreement between model predictions and experimental within the channel may also modulate the kinetics of drug data (Table 2) suggest that the model may be used to unbinding. Alternatively, the possibility exists that the rate of exit of the neutral form of the drug through the channel wall and into the surrounding bilayer could also TABLE 2 Comparison of model predictions with be rate-limiting. experimental data As indicated by Fig. 4, the time course of recovery from 10 M Cocaine 50MM Cocaine cocaine block does not significantly differ over a voltage Experiment* Model$ Experiment* Model$ range of -120 to -150 mV. This suggests that the probability of activation gate opening and/or the lifetime Tonic block 0.94 + 0.06 0.95 0.91 ± 0.04 0.78 of the charged species in the channel over this voltage Steady-statel block at: range are relatively constant. Experimental conditions 5.0 Hz 0.77 ± 0.14 0.75 0.34 ± 0.09 0.33 did not allow discerning the importance of these two 2.5 Hz 0.81 ± 0.14 0.79 0.43 ± 0.09 0.40 mechanisms. 1.25 Hz 0.85 ± 0.13 0.83 0.56 ± 0.05 0.50 1.0 Hz 0.86 ± 0.14 0.85 0.57 ± 0.1 0.53 0.5 Hz 0.83 ± 0.13 0.89 0.68 ± 0.09 0.63 Mathematical modeling Use-dependent block rate (X) at: Previous studies using upstroke velocity measurements in 5.0 Hz 0.05 ± 0.02 0.05 0.12 ± 0.02 0.11 2.5 Hz 0.07 ± 0.02 0.07 0.14 ± 0.02 0.13 cardiac tissue have suggested that the guarded receptor 1.25 Hz 0.12 ± 0.06 0.10 0.18 ± 0.05 0.18 model can adequately describe the blocking effects of 1.0Hz 0.11 ± 0.08 0.12 0.21 ± 0.05 0.20 local anesthetic agents on cardiac Na channels (Starmer 0.5 Hz 0.20 ± 0.14 0.21 0.29 ± 0.03 0.31 and Grant, 1985; Starmer, 1987). Guarded receptor equations have also been recently shown to well describe *Experimental data are mean ± SD (n - 3-6). $Values are model predictions computed from the mean rate constants the on blocking effects of RAC 109 stereoisomers cardiac shown in Table I (n - 10). Na current (Clarkson, 1989). In the present study, equa- Block produced by a train of 30 pulses of 20 ms duration to -20 mV tions based upon the guarded receptor model were found (Vh - -140 mV).

Crumb and Clarkson Cocaine-induced Block of Cardiac Sodium Channels 597 Vr' Na+ o HH+ out V

+ +~~~~~+ Activatio gate

in 0 Drug na ivtio (d - 6-8 A) gate Rested Open Inactivated Rested-blocked

FIGURE 7 An illustration of the guarded receptor model indicating the proposed mechanism by which the activation and inactivation gates regulate binding and unbinding of drug (dark sphere). The model assumes that local anesthetic agents bind to a receptor site in the Na channel for which drug access and egress (indicated by arrows) are regulated by the position of channel gates. There are two pathways for drug access and egress from the channel receptor: a hydrophilic pathway from the cell interior, and a hydrophobic pathway through the lipid bilayer. (Resting block) When channels are held at a hyperpolarized holding potential (e.g., - 140 mV) most channels have their activation gates closed, resulting in guarding of drug access to the receptor over both hydrophilic and hydrophobic pathways, and little net channel block (small tonic block). (Open channel block) When the activation gates open during a depolarizing pulse, drug can bind and unbind rapidly, resulting in measureable channel block observed after one or more briefdepolarizing pulses. (Inactivated channel block) Within 1 ms after a channel opens, it enters another closed (inactivated) state due to closure of an "inactivation" gate. Closure of the inactivation gate prevents drug access and egress over the hydrophilic pathway only. In inactivated channels, lipid soluble drugs have continuous access to the channel receptor via the lipid bilayer pathway, resulting in a slow component of channel block observed with prolonged depolarization. Egress from inactivated channels is accomplished by drug escaping into the lipid bilayer, which for cocaine would involve deprotonation. The illustration is adapted from the molecular model proposed by Guy (1988) depicting a cross section of a Na channel revealing two of the activation gates and the inactivation gate. Positively charged gating segments linked to each activation gate move in response to changes in transmembrane voltage. Outward movement of the gating segments pull the flexible activation gates close to the walls of the channel, permitting ion flow through the channel lumen, as well as unrestricted drug access and egress to the channel receptor. Drug is represented by a dark sphere with a diameter of 6-8 A according to Courtney (1984). Protons (H +) have access and egress pathways (indicated by arrow) to and from the drug binding site as proposed by Woodhull (1973). define rate constants that can adequately predict use- be expected to result in disturbances of cardiac conduc- dependent block of cardiac Na current. tion, and production of conduction block and/or reentrant arrhythmias. Experiments to test this hypothesis in vivo Clinical relevance are now in progress. Cocaine abuse has become a major medical problem in This work was supported by a grant from National Institutes of Health the United States with cocaine related deaths increasing (HL-36096). over the past decade (Mittleman and Wetli, 1984; Pollin, 1985). Case reports suggest a cardiovascular etiology Received for publication 20 July 1989 and in final form 10 (Young and Glauber, 1947; Jonsson et al., 1983; Kos- November 1989. sowsky and Lyon, 1984; Nanji and Filipenko, 1984; Isner et al., 1986) with some implicating conduction block (Young and Glauber, 1947; Jonsson et al., 1983; Nanji REFERENCES and Filipenko, 1984), as the underlying cause. Because cocaine has been shown to be a local anesthetic in nerve, Bean, B. P., C. J. Cohen, and R. W. Tsien. 1983. Lidocaine block of one possible mechanism for cocaine cardiotoxicity is the cardiac sodium channels. J. Gen. Physiol. 81:613-642. alteration of normal conduction within the heart. The Benchimol, A., H. Bartall, and K. B. Desser. 1978. Accelerated ventric- present study has shown that cocaine is a potent cardiac ular rhythm and cocaine abuse. Ann. Intern. Med. 88:519-520. Na channel blocker. A comparison of the recovery time Cahalan, M. D. 1978. Local anesthetic block of sodium channels in constant of cocaine (rT = 8.5 s) with that of lidocaine normal and pronase-treated squid giant axons. Biophys. J. 23:285- (Ts= 1.0 ± 0.4 s) indicates cocaine unbinds approxi- 310. mately eight times slower than lidocaine (Fig. 3 B). Cahalan, M. D., and W. Almers. 1979. Interactions between quaternary However, the of cocaine (7.9 is similar to that of lidocaine, the channel gates, and tetrodotoxin. Biophys. J. 27:39-56. Kdl IuM) Clarkson, C. W. 1989. Stereoselective block of cardiac sodium channels lidocaine (O,u10 M) (Bean et al., 1983). This suggests by RAC109 in single guinea pig ventricular myocytes. Circ. Res. that at equimolar concentrations, cocaine may produce a 65:1306-1323. much greater cumulative block compared with lidocaine Clarkson, C. W., C. H. Follmer, R. E. Ten Eick, L. M. Hondeghem, and due to its slower rate of unbinding. This behavior would J. Z. Yeh. 1988. Evidence for two components of sodium channel

598 Biophysical Journal Volume 57 March 1990 block by lidocaine in isolated cardiac myocytes. Circ. Res. 63:869- Kunze, D. L., A. E. Lacerda, D. L. Wilson, and A. M. Brown. 1985. 878. Cardiac Na currents and the inactivation, reopening, and waiting Courtney, K. R. 1975. Mechanism of frequency-dependent inhibition of properties of single cardiac Na channels. J. Gen. Physiol. 86:691- sodium currents in frog myelinated nerve by the lidocaine derivative 719. GEA 968. J. Pharmacol. Exp. Ther. 195:225-236. Mitra, R., and M. Morad. 1985. A uniform enzymatic method for Courtney, K. R. 1984. Size dependent kinetics associated with drug dissociation of myocytes from hearts and stomachs of vertebrates. block of sodium current. Biophys. J. 45:42-44. Am. J. Physiol. 249:H1056-H1060. Crumb Jr., W. J., and C. W. Clarkson. 1989. Guarded receptor analysis Mittleman, R. E., and C. V. Wetli. 1984. Death caused by recreational of cocaine interaction with cardiac sodium channels. Biophys. J. cocaine use: an update. JAMA (J. Am. Med. Assoc.). 252:1889- 55:314 (Abst.) 1893. Finkle, B. S., and K. L. McCloskey. 1977. The forensic toxicology of Nanji, A. A., and J. D. Filipenko. 1984. Asystole and ventricular cocaine. In National Institute of Drug Abuse Research Monograph fibrillation associated with cocaine intoxication. Chest. 85:132-133. 13. R. C. Petersen and R. C. Stillman, editors. National Institute of Pollin, W. 1985. The danger of cocaine. JAMA (J. Am. Med. Assoc.). Drug Abuse, Rockville, MD. 7-10. 254:98. Follmer, C. F., R. E. Ten Eick, and J. Z. Yeh. 1987. Sodium current Schwarz, W., P. T. Palade, and B. Hille. 1977. Local anesthetics: effect kinetics in cat atrial myocytes. J. Physiol. 384:169-197. of pH on use-dependent block of sodium channels in frog muscle. Fozzard, H. A., D. A. Hanck, J. C. Makielski, B. E. Scanley, and M. F. Biophys. J. 20:343-368. Sheets. 1987. Sodium channels in cardiac Purkinge cells. Experien- Spiehler, V. R., and D. Reed. 1985. Brain concentrations of cocaine and tia. (Base 1). 43:1162-1168. benzoylecgonine in fatal cases. J. Forensic Sci. 30:1003-1011. Gintant, G. A., and B. F. Hoffman. 1984. Use-dependent block of Starmer, C. F. 1987. Theoretical characterization of ion channel cardiac sodium channels by quaternary derivatives of lidocaine. blockade. Competitive binding to periodically accessible receptors. Pfluegers Arch. Eur. J. Physiol. 400:121-129. Biophys. J. 52:405-412. Guy, R. H. 1988. A model relating the structure of the sodium channel Starmer, C. F., and A. 0. Grant. 1985. Phasic ion channel blockade: a to its function. In Molecular Biology of Ionic Channels. J. Hoffman kinetic model and parameter estimation procedure. Mol. Pharmacol. and G. Giebisch, editors. Harcourt Brace Jovanovich, New York. 28:348-356. 289-308. Starmer, C. F., and K. R. Courtney. 1986. Modeling ion channel 0. A. E. Hamill, P., Marty, Neher, B. Sakmann, and F. J. Sigworth. blockade at guarded binding sites: application to tertiary drugs. Am. 1981. Improved patch-clamp techniques for high-resolution current J. Physiol. 251:H848-H856. recording from cells and cell-free membrane patches. Pfluegers Arch. Eur. J. Physiol. 391:85-100. Starmer, C. F., A. 0. Grant, and H. C. Strauss. 1984. Mechanisms of Hille, B. 1977. Local anesthetics: hydrophilic and hydrophobic path- use-dependent block of sodium channels in excitable membranes by ways for the drug-receptor reaction. J. Gen. Physiol. 69:497-515. local anesthetics. Biophys. J. 46:15-27. Hondeghem, L. M., and B. G. Katzung. 1977. Time- and voltage- Starmer, C. F., J. Z. Yeh, and J. Tanguy. 1986. A quantitative dependent interactions of antiarrhythmic drugs with cardiac sodium description of QX222 blockade of sodium channels in squid axons. channels. Biochim. Biophys. Acta. 472:373-398. Biophys. J. 49:913-920. Hondeghem, L. M., and B. G. Katzung. 1980. Test of a model of Strichartz, G. R. 1973. The inhibition of sodium current in myelinated nerve antiarrhythmic drug action. Effect of quinidine and lidocaine on by quaternary derivatives of lidocaine. J. Gen. Physiol. 62:37- myocardial conduction. Circulation. 61:1217-1224. 57. Isner, B. S., N. A. M. Estes, P. D. Thompson, M. R. Castarzo-Nordin, Wang, G. K. 1988. Cocaine-induced closures of single batrachotoxin- R. Subramanian, G. Miller, G. Kastas, K. Sweeney, and W. Q. activated Na+ channels in planar lipid bilayers. J. Gen. Physiol. Sturner. 1986. Acute cardiac events temporally related to cocaine 92:747-765. abuse. N. Engi. J. Med. 315:1438-1444. Wetli, C. V., and R. K. Wright. 1979. Death caused by recreational Jonsson, S., M. O'Meara, and J. B. Young. 1983. Acute cocaine cocaine use. JAMA (J. Am. Med. Assoc.). 241:2519-2521. poisoning. Am. J. Med. 75:1061-1064. Woodhull, A. M. 1973. Ionic blockade of sodium channels in nerve. J. Kojima, M., and T. Ban. 1988. Nicorandil shortens action potential Gen. Physiol. 61:687-708. duration and antagonises the reduction of V , by lidocaine but not by Yeh, J. Z., and J. Tanguy. 1985. Na channel activation gate modulates disopyramide in guinea-pig papillary muscles. Naunyn-Schmiede- slow recovery from use-dependent block by local anesthetics in squid berg's Arch. Pharmacol. 337:203-212. giant axons. Biophys. J. 47:685-694. Kossowsky, Wa., and A. F. Lyon. 1984. Cocaine and acute myocardial Young, D., and J. J. Glauber. 1947. Electrocardiographic changes infarction. A probable connection. Chest. 86:729-731. resulting from acute cocaine intoxication. Am. Heart J. 34:272-279.

Crumb and Clarkson Cocaine-induced Block of Cardiac Sodium Channels 599